I like context when it comes to talking hockey. A lot.
And think about it; you do too. We have certain statistical markers that stick in everyone's mind, and they all are barometers (read: context) for a player's performance. 50 goals is a damn good year, no? A point-per-game is solid, too. These days, a goaltender's GAA better not be north of 2.75, and his save percentage better not be south of .905. All of these markers were ingrained at some point; 50 goals likely around the time of the "50 goals in 50 games" craze of the Richard days, a point-per-game since forever, and those goalie numbers only since the early 1990s. Yet they generally have the same kind of clout that a .300 batting average has in baseball, or 4 yards a carry or 60 percent completion in football, or 45 percent shooting in basketball. They all recognize the average, and thus help demarcate above-average performance. In other words, they help make a metric intuitive for more people.
And I put it on advanced statisticians (I'm not claiming to be one, as you'll see below I'm not doing advanced stuff) to work towards ingraining those league averages and context. It's within everyone's intuition that above 50% on faceoff wins is good, because we know league average is 50%. But what's league average Corsi? What's league average Corsi at X percentage Zone Starts? What is a player's expected zone shift (offensive zone start minus offensive zone finish)? What's a tough QualComp/QualTeam on this team, or that team? Advanced stats cannot dream of entering the average fan's lexicon without these kinds of questions answered.
I've been admiring Eric T.'s work at Broad Street Hockey from afar, as he's been working on those kinds of questions with Corsi. Have a look here:
- Balanced Corsi - creating an expected Corsi metric for each situation (be sure to read the comments)
- Balanced Corsi Data Dump - the previous study reduced the player population to those who'd played a large amount of the team's games; this includes the rest
I also made similar attempts with Zone Starts, creating a metric (Balanced Zone Shift, or BZS):
- Expected Zone Shift, 2007-08 to 2010-11 - the initial post creating a chart and graph
- Balanced Zone Shift (BZS) Single Season Leaders, 2007-08 to 2010-11 - self-explanatory
- Balanced Zone Shift Leaders and Cabooses, 2007-08 to 2010-11 - self-explanatory
- BZS 2.0: Ironing Things Out a Bit - the previous posts only worked with half of the 2010-11 season, so I wanted to do one after the end of 2010-11 and feature a refined chart and equation for the stats folks to work with
I'm sure there are more things out there, and I'd encourage anybody that can link to similar posts to mention them in the comments. I'll do a second "Context" post and include those.
Now, for what I did...I took 2010-11's data, and slapped together a bunch of league-averages, splitting forwards and defencemen and taking out players that had played fewer than 20 games and 10 minutes per game. That way, we don't have anything fudging our rate statistics, even slightly. I find it very important to note the differences between forwards and defencemen in most of these categories. Fair warning: I rounded to the whole number for the counting stats, but not for the per game stats.
First off, our average forward played 66 games. His rates and numbers, put to an 82-game season, were as follows:
- 12.85 even-strength minutes per game - 1.83 powerplay minutes - 1.05 shorthanded minutes - 15.74 total minutes
- 166 shots, 60 missed shots, 18 goals, 10.62 shooting percentage, 99 hits, 36 blocked shots, 39 takeaways, and 31 giveaways
Our average defenceman played 62 games:
- 16.42 even-strength minutes per game - 1.96 powerplay minutes - 1.57 shorthanded minutes - 19.96 total minutes
- 108 shots, 51 missed shots, 5 goals, 4.84 shooting percentage, 111 hits, 126 blocked shots, 27 takeaways, and 44 giveaways
Some big differences in these categories, especially in time on-ice, shots, shooting percentage, and blocked shots. The rate statistics are trickier, because just going per-game doesn't really help the guys who play only 10 minutes versus ones playing 20; so instead, I gave forwards their rates per 15 minutes and defencemen per 20 minutes. Why did I do that? As you can see, these revolve closer to the average amount of time they play in a game.
Our average forward, in his 15 minutes, recorded:
- 1.93 shots, 1.15 hits, 0.42 blocked shots, 0.45 takeaways, and 0.36 giveaways
Our average defenceman, in his 20 minutes, recorded:
- 1.32 shots, 1.35 hits, 1.54 blocked shots, 0.33 takeaways, and 0.54 giveaways
Even in these data, context is important, and that's something I'm going to work on for the next post. There are strong relationships between some of these categories and the kinds of playing time a player receives; we can look at blocked shots and their relationship to increased time on the penalty kill, or see if hits have anything to do with, well, anything. What is worth laying out, but is pretty much undeniable, is that playing time has a profound effect on most of those old markers (50 goals, point-per-game) we've grown accustomed to, in addition to the newer in vogue counting stats.